--- title: GPT-5.6 Sol runs at 750 tokens per second on Cerebras... description: OpenAI's GPT-5.6 Sol is running on Cerebras hardware at up to 750 tokens per second, a dramatic increase from GPT-5.5 XHigh's 70-100 TPS. Sam Altman noted 5.6... canonical: https://daily.dev/posts/gpt-5-6-sol-runs-at-750-tokens-per-second-on-cerebras-hardware-wgmjcpgt4 twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: GPT-5.6 Sol runs at 750 tokens per second on Cerebras hardware | daily.dev og:description: OpenAI's GPT-5.6 Sol is running on Cerebras hardware at up to 750 tokens per second, a dramatic increase from GPT-5.5 XHigh's 70-100 TPS. Sam Altman noted 5.6... og:url: https://daily.dev/posts/gpt-5-6-sol-runs-at-750-tokens-per-second-on-cerebras-hardware-wgmjcpgt4 og:image: https://api.daily.dev/og/posts/WgmJCPGT4.png og:image:alt: GPT-5.6 Sol runs at 750 tokens per second on Cerebras hardware og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # GPT-5.6 Sol runs at 750 tokens per second on Cerebras hardware **[Collections](https://daily.dev/sources/collections)** · 2 min read · 6 upvotes · 3 comments ## Summary OpenAI's GPT-5.6 Sol is running on Cerebras hardware at up to 750 tokens per second, a dramatic increase from GPT-5.5 XHigh's 70-100 TPS. Sam Altman noted 5.6 Sol is also 54% more efficient on agentic coding tasks. The speed gain is particularly meaningful for agentic workloads, where subagents handle planning, iteration, and tool coordination — each loop requiring inference, making raw throughput a real bottleneck. Cerebras achieves this through its wafer-scale architecture, which keeps large amounts of SRAM on-chip, avoiding the memory bandwidth wall that GPU-based systems face when shuttling data to external HBM. This deployment signals that inference speed is becoming infrastructure-critical as AI shifts toward agentic systems in coding, research, voice, and robotics. Whether Cerebras can expand beyond this partnership to broader hyperscaler adoption remains the key open question. ## Content GPT-5.6 Sol is running on Cerebras at up to 750 tokens per second. For context, GPT-5.5 XHigh was doing 70–100 TPS. That's a significant jump. Sam Altman also noted that GPT-5.6 Sol is 54% more efficient on agentic coding tasks compared to its predecessor. Early anecdotal results back that up — developer antirez reported that Sol wrote a Laguna S2.1 inference implementation almost unassisted, generating at 50 t/s with fast prefill, following the architecture of two other models in the DwarfStar project. ## Why inference speed matters for agents The speed gap between Cerebras and GPU-based inference comes down to memory architecture. Standard GPU inference hits a memory bandwidth wall because models have to constantly shuttle weights between HBM (high-bandwidth memory) and compute. Cerebras sidesteps this with its wafer-scale chip, which integrates a large SRAM pool directly on-chip. The company claims this makes inference up to 15x faster than GPUs for certain workloads. This matters more as models like Sol use subagents internally — breaking complex tasks into planning, iteration, and tool calls that each require their own inference passes. More autonomous agent loops means more inference, and latency compounds quickly. At 750 TPS, those loops run fast enough to feel nearly instantaneous. ## The investment angle Cerebras ($CBRS) is getting attention from investors who think the memory bottleneck story is underappreciated. The common trade has been HBM suppliers like SK Hynix and Micron, but Cerebras's approach is architecturally different — it manufactures its inference memory directly into the wafer rather than relying on external HBM at all. OpenAI's decision to run GPT-5.6 Sol on Cerebras is the kind of validation that could attract more hyperscaler interest. Whether that translates to an AWS partnership or another major adoption in the next 12–24 months remains to be seen, but the technical case for high-speed on-chip inference in an agentic world is getting harder to ignore. ## Community discussion Top comments from developers on daily.dev. **@starfallcodes** · 1 upvotes > ngl its so cool **@luispsarmiento** · 1 upvotes > I know now something new: tokens per second. --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#openai](https://daily.dev/tags/openai), [#ai-inference](https://daily.dev/tags/ai-inference) [View this post on daily.dev](https://daily.dev/posts/gpt-5-6-sol-runs-at-750-tokens-per-second-on-cerebras-hardware-wgmjcpgt4) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"GPT-5.6 Sol runs at 750 tokens per second on Cerebras hardware","url":"https://daily.dev/posts/gpt-5-6-sol-runs-at-750-tokens-per-second-on-cerebras-hardware-wgmjcpgt4","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/gpt-5-6-sol-runs-at-750-tokens-per-second-on-cerebras-hardware-wgmjcpgt4"},"datePublished":"2026-07-17T10:45:52.774Z","dateModified":"2026-07-22T22:33:24.097Z","description":"OpenAI's GPT-5.6 Sol is running on Cerebras hardware at up to 750 tokens per second, a dramatic increase from GPT-5.5 XHigh's 70-100 TPS. Sam Altman noted 5.6...","image":"https://pbs.twimg.com/media/HNQ_fRmWcAAvOYw.jpg","thumbnailUrl":"https://pbs.twimg.com/media/HNQ_fRmWcAAvOYw.jpg","isAccessibleForFree":true,"articleSection":"Collections","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Collections","logo":"https://media.daily.dev/image/upload/s--fk_6ycEi--/f_auto,q_auto/v1780996001/logos/collections?_a=BAMAMiWQ0","url":"https://daily.dev/sources/collections"},"commentCount":3,"discussionUrl":"https://daily.dev/posts/gpt-5-6-sol-runs-at-750-tokens-per-second-on-cerebras-hardware-wgmjcpgt4","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":6},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":3}],"keywords":"ai-agents,openai,ai-inference","timeRequired":"PT2M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Collections","item":"https://daily.dev/sources/collections"},{"@type":"ListItem","position":3,"name":"GPT-5.6 Sol runs at 750 tokens per second on Cerebras hardware"}]} {"@context":"https://schema.org","@type":"WebPage","@id":"https://daily.dev/posts/gpt-5-6-sol-runs-at-750-tokens-per-second-on-cerebras-hardware-wgmjcpgt4","comment":[{"@type":"Comment","text":"ngl its so cool","datePublished":"2026-07-17T15:59:07.956Z","url":"https://daily.dev/posts/WgmJCPGT4#c-nUFAiPCm6","author":{"@type":"Person","name":"Starfall :D","url":"https://daily.dev/starfallcodes","image":"https://media.daily.dev/image/upload/s--4xoTqFbk--/f_auto/v1786806197/avatars/avatar_ecPDW57nUVRidLxwsxruz?_a=BAMAMicg0"},"interactionStatistic":{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1}},{"@type":"Comment","text":"I know now something new: tokens per second.","datePublished":"2026-07-17T16:26:16.864Z","url":"https://daily.dev/posts/WgmJCPGT4#c-lRzoegu0E","author":{"@type":"Person","name":"Luis Puc","url":"https://daily.dev/luispsarmiento","image":"https://avatars.githubusercontent.com/u/25454970?v=4"},"interactionStatistic":{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1}}]} ```